Emergence, Locating and Measuring It - Systematic Design for Emergence in Cellular Nonlinear Networks - page 54

computation is the dynamic evolution characterized by glider interaction, and
there is a specified way to interpret the result of the computation as a terminal
state (obtained after a certain number of iterations).
There are also emergent phenomena which do not rely on gliders but still may
have some meaningful computational properties. Such phenomena are better re-
lated to natural computation of the type seen in the brain, in social collectivities,
etc. As shown in the applicative chapters of this topic, such emergent behaviors
can successfully be used to certain perceptual-related tasks such as feature extrac-
tion, image compression or time-sequence classification with the advantage of re-
ducing the implementation complexity when compared to traditional, algorithmic
solutions for similar tasks.
Detecting the presence of interacting gliders or other emergent phenomena by
visual observation of a CA simulation is simple in principle. But for families of
CA with large numbers of individuals (e.g. “1s9”, “2s9” or “1a9”, “2a9”) visual
detection becomes completely unpractical. Therefore we need to develop meas-
ures of emergence to evaluate and quantify the variety of phenomena mentioned
above, including the likelihood of emergent gliders.
Wolfram defines Class IV as the one containing the most interesting behaviors,
mostly based on gliders. Based on a visual observation of all 1,024 CAs from the
1s5 family, we selected the next three examples shown in Fig. 4.5 as representa-
tive.

(a) (b) (c)

Fig. 4.5. Different behaviors of CAs belonging to Wolfram's Class IV

Again, like in the case of ID = 133 (Fig. 4.3b), if the CA with ID=684, dis-
played above in Fig. 4.5b), is simulated for a longer period, it turns out that the

behavior actually belongs to Class II, i.e. ending into a low periodic pattern

with a high degree of order. This is clearly visible in Fig. 4.6, and confirms that
there is a strong connection between the presence of interacting gliders, long tran-
sients (even if the simulation takes a small number of iterations) and a “halting”

process (described by the low periodic pattern reached after a large number of

it erations) specific for any computational process.